SPAM with lots of random words and good words
Boris 'pi' Piwinger
3.14 at logic.univie.ac.at
Mon Jan 19 11:24:13 CET 2004
Manvendra Bhangui wrote:
> I have been seeing few messages like the one forwarded below.
Why do you send spam to this list? Just to spoil databases?
> These mails have lot of good words. These get past bogofilter the first time.
Not at all. You just managed to produce the first false
positive in ages here. Obviously your message contains
incredible amounts of spammish words:
> X-Bogosity: Spam, spamicity=0.500, version=0.16.3
> int cnt prob spamicity histogram
> 0.00 52 0.039397 0.017582 ##########################
> 0.10 33 0.154138 0.052105 #################
> 0.20 0 0.000000 0.052105
> 0.30 0 0.000000 0.052105
> 0.40 0 0.000000 0.052105
> 0.50 0 0.000000 0.052105
> 0.60 0 0.000000 0.052105
> 0.70 0 0.000000 0.052105
> 0.80 2 0.852666 0.075560 #
> 0.90 97 0.960114 0.526999 ################################################
> Is it advisable to increase the value of robx to something closer
> to the spam cutoff value so that words that do not occur in wordlist.db
> get a higher score.
There are different choices:
1) Use bogotune to find better parameters.
2) Completely change your training method, see:
http://cvs.sourceforge.net/viewcvs.py/*checkout*/bogofilter/bogofilter/doc/bogofilter-faq.html?rev=HEAD&content-type=text/html#training
pi
More information about the Bogofilter
mailing list